1
CVPR 2026最热方向，被一家杭州团队率先跑进了端侧！
CVPR 2026热门方向，杭州团队率先在端侧实现精准定位与实时运动规划，不再“猜坐标”而是选区域。
VLM- R1之后再次出手！全球首个端侧流式多模态来了！
2
CodePercept: Code-Grounded Visual STEM Perception for MLLMs
针对MLLMs视觉STEM感知难题，CodePercept创新引入代码接地机制，为多模态理解开辟新路径，已被CVPR2026收录。
arXiv:2603.10757v2 Announce Type: replace Abstract: When MLLMs fail at Science, Technology, Engineering, and Mathematics (STEM) visual reasoning, a fu…
3
SAMTok: Representing Any Mask with Two Words
用两个词就能表达任意分割掩码，SAM最新变体实现高效语义token化表示。
arXiv:2601.16093v2 Announce Type: replace Abstract: Pixel-wise capabilities are essential for building interactive intelligent systems. However, pixel…
4
OccAny: Generalized Unconstrained Urban 3D Occupancy
被CVPR 2026接收的3D occupancy通用框架，无需严格约束即可应对复杂城市场景，为自动驾驶感知提供新思路。
arXiv:2603.23502v2 Announce Type: replace Abstract: Relying on in-domain annotations and precise sensor-rig priors, existing 3D occupancy prediction m…
5
Differences in Detection: Explainability Where it Matters
探索目标检测模型的可解释性，聚焦关键区域差异，CVPR 2026 Workshop论文。
arXiv:2606.07503v1 Announce Type: new Abstract: We propose Differences in Detection (DnD), an intuitive method to compare two object detection models.…
6
MotionEnhancer: Leveraging Video Diffusion for Motion-Enhanced Vision-Language Models
让视觉语言模型看懂动作——利用视频扩散模型注入运动信息，提升动态场景理解能力，CVPR 2026接收。
arXiv:2606.06853v1 Announce Type: cross Abstract: The new era has witnessed a remarkable capability to extend Vision-Language Models (VLMs) for tackli…
7
Attention Consistent Longitudinal Medical Visual Question Answering Guided by Vision Foundation Models
用视觉基础模型引导注意力一致性，解决纵向医学图像问答难题，CVPR 2026 workshop入选研究。
arXiv:2606.06534v1 Announce Type: cross Abstract: Longitudinal medical visual question answering (VQA) requires reasoning about anatomical differences…
8
Answer Self-Consistency with Margin-Triggered Question Re-Arbitration for the CVPR 2026 VidLLMs Challenge
CVPR 2026视频大模型挑战赛新方案：用边际触发问题重仲裁提升回答自一致性，精准优化视频问答可靠性。
arXiv:2606.04323v1 Announce Type: new Abstract: In this report, we present our solution for Track 2 of the CVPR 2026 VidLLMs Challenge. This track eva…
9
CVPR 2026，英伟达特斯拉Waymo一块听中国公司讲物理AI
CVPR 2026上，英伟达、特斯拉、Waymo齐聚，听中国公司小鹏分享物理AI新思路，世界模型与VLA路线碰撞。
率先实现物理AI闭环飞轮
10
Efficient and Training-Free Single-Image Diffusion Models
提出一种无需额外训练的扩散模型，仅需单张图像即可生成高质量样本，CVPR 2026论文。
arXiv:2606.04299v1 Announce Type: new Abstract: We consider the problem of generating images whose internal structure -- defined by the distribution o…
11
Eliciting Complex Spatial Reasoning in MLLMs through Wide-Baseline Matching
CVPR 2026 新作：用宽基线匹配技术激发多模态大模型的空间推理潜力，突破复杂场景理解瓶颈。
arXiv:2606.03577v1 Announce Type: new Abstract: Wide-baseline matching (WBM) requires integrating geometric understanding, viewpoint changes, fine-gra…
12
Training-Free Composed Video Retrieval via Visual Representation-Guided Video-LLM Reasoning
无需额外训练，利用视觉表示引导视频大模型推理，实现组合视频检索，CVPR 2026 workshop 最新成果。
arXiv:2606.02321v1 Announce Type: new Abstract: Recent advances in large vision-language models have expanded video retrieval from simple text-based s…
13
EarlyTom: Early Token Compression Completes Fast Video Understanding
CVPR 2026最新研究：通过早期Token压缩实现快速视频理解，颠覆传统视频处理效率瓶颈。
arXiv:2605.30010v1 Announce Type: new Abstract: Video large language models (Video-LLMs) have demonstrated strong capabilities in video understanding …
14
BitTP: The Lightweight Trajectory Prediction Model with BitLLM for Edge-Devices
用比特级LLM做轨迹预测，轻量到能塞进边缘设备，CVPR 2026论文。
arXiv:2605.29705v1 Announce Type: new Abstract: Trajectory prediction is a fundamental task for autonomous systems, requiring complex reasoning about …
15
EpiAgent: An Agent-Centric System for Ancient Inscription Restoration
CVPR 2026录用论文，提出智能体驱动的古代铭文恢复系统，实现精准的破损文字重建。
arXiv:2604.09367v2 Announce Type: replace Abstract: Ancient inscriptions, as repositories of cultural memory, have suffered from centuries of environm…
16
CuriosAI Submission to the CASTLE Challenge at EgoVis 2026
CVPR EgoVis研讨会CASTLE挑战第4名方案，揭秘AI视觉竞赛中的前沿技术。
arXiv:2605.27800v1 Announce Type: new Abstract: CASTLE 2026 asks 185 multiple-choice questions over 600+ hours of synchronized multi-view egocentric v…
17
Model Merging on Loss Landscape: A Geometry Perspective
从几何视角理解损失景观中的模型合并，CVPR最新研究揭示关键规律。
arXiv:2605.26693v1 Announce Type: cross Abstract: Model merging offers a promising avenue for knowledge integration and parallel development without r…
18
Multi-SpatialMLLM: Multi-Frame Spatial Understanding with Multi-Modal Large Language Models
CVPR 2026论文，提出多帧空间理解新方法，刷新多模态大模型空间推理能力。
arXiv:2505.17015v2 Announce Type: replace-cross Abstract: Multi-modal large language models (MLLMs) have rapidly advanced in visual tasks, yet their s…
19
Scene Reconstruction as Mapping Priors for 3D Detection
用场景重建生成地图先验，大幅提升3D目标检测精度，CVPR 2026前沿论文。
arXiv:2605.22997v1 Announce Type: new Abstract: In autonomous driving, mapping is critical for motion planning but remains an under-utilized resource …
20
MagicFuse: Single Image Fusion for Visual and Semantic Reinforcement
CVPR 2026新作，提出一种单图像融合方法，同时增强视觉细节与语义信息，为图像处理开辟新路径。
arXiv:2602.01760v2 Announce Type: replace Abstract: This paper focuses on a highly practical scenario: how to continue benefiting from the advantages …